LiDAR Point Cloud Object Classification

LiDAR point cloud data serves as an machine vision alternative other than image. Its advantages when compared to image and video includes depth estimation and distance measruement. Low-density LiDAR point cloud data can be used to achieve navigation, obstacle detection and obstacle avoidance for mobile robots. autonomous vehicle and drones. In this metadata, we scanned over 1200 objects and classified it into 4 groups of object namely, human, cars, motorcyclist. We also recorded the objects orientation heading for single image orientation detection which enables movement forecast and estimation.

激光雷达点云数据作为一种机器视觉的替代方案，超越了传统图像。相较于图像与视频，其优势在于能够实现深度估计与距离测量。低密度激光雷达点云数据可应用于移动机器人、自动驾驶汽车与无人机等领域的导航、障碍物检测与避障。在本元数据中，我们对超过1200个物体进行了扫描，并对其进行分类，分为四组对象，即人类、汽车、摩托车。此外，我们还记录了对象的朝向航向信息，以实现单图像朝向检测，从而促进运动预测与估算。